I’ve been experimenting with AI coding tools a lot recently.
They’re great for getting something working fast.
But I’ve noticed a pattern:
Things look clean.
They run fine.
But I don’t always fully understand everything that’s happening anymore.
Especially when it starts touching:
APIs
data
permissions
or multiple layers of the system
For MVPs, that’s fine.
But for production, it feels different.
Some people here mentioned they:
only use AI for small scoped tasks
review everything line by line
or spend more time in planning before generating anything
That feels like the right direction.
Because the issue doesn’t seem to be “AI writes bad code.”
It’s more:
→ how easily we start trusting code once it works
→ and how that affects review depth
I’m curious how others are handling this.
Do you:
treat AI as a helper you fully control
or let it generate larger parts of the system
And more importantly:
Do you trust AI-generated code enough to push it to production?
No responses yet.